【作者】 李云刚；

【导师】 毕晓君；

【作者基本信息】 哈尔滨工程大学 ， 信号与信息处理， 2010， 硕士

【摘要】 自由搜索算法是近年来提出的一种较新的群体智能优化算法,其算法自身还有很大的研究空间,因此对自由搜索算法进行有效的改进仍然是目前学者研究的一个热点。本文提出一种基于灾变策略的自由搜索改进算法。该算法在群体进行下一轮搜索之前进行一次灾变判断,符合灾变条件的个体将进行相应的灾变操作,将重新初始化下一轮搜索起点。从自然界规律上说该算法较基本的自由搜索算法具有更优秀的全局搜索能力。将本文算法和现有的两种性能最好的自由搜索改进算法(自适应搜索算法和基于粗细粒交叉的自由搜索算法)进行仿真对比,通过对三个基准函数的优化测试,仿真结果表明本文提出的算法在收敛精度和收敛速度上都要优于现有的两种改进算法。基于最大熵的阈值法图像分割的关键是找到一组合适的阈值向量,使得分割后图像的总熵最大,而总熵恰恰是一个以阈值为自变量的多维函数,在以往的阈值搜索中,存在精度不高的缺陷,鉴于自由搜索算法具有强大的搜索能力,本文提出了基于改进自由搜索算法(即基于灾变策略的自由搜索算法)的最大熵阈值图像分割算法,通过比较本文算法和目前在基于最大熵的阈值法图像分割中应用较好的改进粒子群算法的仿真结果,本文提出的算法可以获得更大的熵值和统一度量标准u值,验证了该改进算法在图像分割中应用的有效性,拓宽了自由搜索算法的应用领域。更多还原

【Abstract】 Free Search (FS) is a new swarm intelligent algorithm put forward few years ago. There is wider research space for algorithm itself. So proposing efficient improvement of FS is still a hot topic among the researchers now. A novel improved free search algorithm is proposed in this paper, which is based disaster policy. In this algorithm, the individual will make a judge of the disaster, the ones who match the condition will be initialized starting point of the next research cycle. According to the natural, the new algorithm has more powerful global searching ability. The simulation results showed that, compared with other improved FS algorithms(adaptation search algorithm and the FS based cross operator) proposed before, it is improved virtually on convergence precision and speed by using the FS algorithm based disaster policy to optimize 3 typical benchmarks.Finding a suitable combination of thresholds is the key to threshold image segmentation based on maximum entropy. The entropy is the function of the threshold, it couldn’t reach precision high enough by other optimization algorithm. In this paper, Free Search was proposed to improve the method to serve the problem. By the experiment of the Lena and Pepper image, The simulation results showed that, the improved method can select optimal threshold, which can achieve more perfect segmentation effect. It is effective to use the improved algorithm in the image segmentation, whose application field is widened.更多还原